Hands-On Exercise: Model Explainability CT-AI Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   What is the primary output of a SHAP (SHapley Additive exPlanations) analysis for a single prediction?

   • AA decision tree that mimics the model's logic for the entire dataset.
   • BA set of feature contribution values that sum to the difference between the prediction and the baseline.
   • CA single global importance score for each feature across all predictions.
   • DA counterfactual example that changes the prediction to a desired outcome.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Which statement best describes the purpose of LIME (Local Interpretable Model-agnostic Explanations)?

   • AIt visualizes the internal neurons of a deep neural network.
   • BIt computes exact Shapley values for every feature in the model.
   • CIt explains individual predictions by approximating the model locally with an interpretable model.
   • DIt provides a global ranking of feature importance across the entire dataset.
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A model-agnostic explanation tool is being used to explain a text-classification model. The tool requires the user to specify the number of features to show in the explanation. The QA team wants to ensure the explanation is both interpretable and sufficiently detailed for a non-technical stakeholder. What is the most appropriate consideration when setting this parameter?

   • AAlways use the minimum number of features to keep the explanation simple.
   • BAlways use the maximum number of features to ensure the explanation is complete.
   • CBalance the number of features so the explanation is concise enough to be understood but includes the most influential features.
   • DThe number of features does not affect the quality of the explanation.
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   A tester wants to explain why a specific loan application was rejected by a complex ensemble model. Which explainability method is most appropriate for this task?

   • AGlobal feature importance from a random forest
   • BLIME
   • Ct-SNE visualization
   • DPartial dependence plot
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A team is testing an image-classification model for a medical diagnosis tool. They use a saliency map to explain a prediction. The saliency map highlights pixels in the background of the image as highly influential. What should the team conclude about the quality of this explanation?

   • AThe explanation is correct because saliency maps always highlight the most important pixels.
   • BThe explanation is high quality because it shows the model is considering the entire image.
   • CThe explanation may be of low quality because it highlights irrelevant regions, indicating the model may be relying on spurious correlations.
   • DThe team should ignore the saliency map because image explanations are unreliable.
   Reveal Solution Discussion Select an answer first